ai-platform/roadmap
All pages

Start here

  1. How to use this roadmap
  2. Two kinds of role
  3. Read code, write it rarely
  4. Target profile
  5. Eight principles
  6. Work on a real product

Phases

  1. 01Direct coding agents and read their code
  2. 02Learn how models behave and what they cost
  3. 03Build retrieval that finds the right evidence
  4. 04Build tool servers with MCP
  5. 05Design agent workflows that survive failure
  6. 06Run agent workloads on AWS
  7. 07Trace every step an agent takes
  8. 08Build evals that catch regressions
  9. 09Run agents reliably in production
  10. 10Secure agents and the code they write
  11. 11Build an internal AI platform
  12. 12Get teams to adopt the platform
  13. 13Build Forge, the capstone platform
  14. Elective: Kubernetes

After the phases

  1. Staff skills
  2. 12-month plan
  3. A typical week
  4. Show your work
  5. Tools to know
  6. What to skip
  7. Final checklist

Phase 6 · Months 5 to 7

Run agent workloads on AWS

Why cloud basics matter here

AI platform projects often fail on cloud basics: IAM policies that allow too much, NAT gateway costs nobody noticed, retries that make an outage worse, Lambda concurrency without a limit, and no separation between the control plane and the workload. Agents will write infrastructure code for you. They will not warn you that a policy allows too much unless you know to ask.

Compute

Learn Lambda in depth: concurrency, cold starts, SnapStart, packaging, and failure destinations. Learn Amazon Elastic Container Service (ECS) on Fargate for container workers, and enough EC2 to understand what runs underneath. Know when serverless fits and when a pool of long-running workers is cheaper or simpler.

Messaging and workflows

Learn SQS, SNS, EventBridge, and Step Functions. Learn visibility timeouts, dead-letter queues, retry behaviour, ordering, deduplication, and at-least-once delivery. Each of these maps directly to a way of running agents reliably.

Storage, networking, and identity

Learn when to use S3, DynamoDB, RDS for Postgres, and ElastiCache. Learn VPCs, subnets, security groups, NAT, private endpoints, and load balancers. Learn IAM, STS, KMS, and Secrets Manager in depth. You should be able to explain exactly which principal can do what, and why.

Bedrock and managed agent services

Learn how to call models through Bedrock and how its provider abstraction works. Learn the managed agent services AWS offers, such as Bedrock AgentCore. Keep your own agent interfaces portable, so that you can use managed services without tying every part of the architecture to them.

Build an event-driven agent runner

Take agent jobs from an API through a durable queue. Workers lease jobs, save state, emit traces, and send jobs that keep failing to a dead-letter queue. Deploy with Terraform or CDK, keep environments separate, give each role the least privilege it needs, and estimate what the infrastructure costs. Review every IAM policy that the agent writes yourself.

You are done when

You can draw an AWS architecture for an agent platform on a whiteboard, explain its network and IAM boundaries, predict how it fails, and deploy it from code without using the console.

Exit checklist

0 of 6 done